From: Robert J. Bradbury (bradbury@aeiveos.com)
Date: Sun Aug 18 2002 - 12:43:58 MDT
On Sun, 18 Aug 2002, Anders Sandberg, commenting on my comments wrote:
> I'm no epidemologist, but isn't this a somewhat optimistic assumption?
Isolation isn't the hard part at least if its based on anything close
to organisms we know about. All one has to do is use probes for common
or conserved parts (e.g. ribosomal RNA, tRNA, catalytic sites, etc.).
You use a machine like this:
http://www.nanosphere-inc.com/technology/detectionsystem.htm
You test for everything at once. There are other machines on the market
with similar capabilities -- it is a very hot area for funding right now.
The DoD is trying to get soldiers informed of any exposure to a bioweapon
to within 20 minutes in the field.
Once you have a reasonable match you can use that to pull out the entire
chromosome and shotgun sequence or primer-walk sequence the entire thing.
> People start to get sick, and the carrier is caught within a few days. I
> saw a documentary yesterday about a cryptosporidium outbreak, and despite
> the pathogen being known in the literature it still took quite a while
> for the authorities to even identify it.
Thats because we don't have a catalog for all subspecies yet and machines
such as that pointed out above aren't in common use yet. But a well
equipped molecular biology lab setup with a PCR machine and the right
sets of primers should be able to identify any known pathogen within
a day.
> What about adding the IL-4 gene as in the Australian mousepox
> experiments? This is a gene which occurs naturally, and might
> be disregarded as a normal gene despite making at least the mousepox
> virus very nasty.
Obviously one shouldn't be finding the IL-4 gene in a mousepox genome.
If one does thats a for sure signature you are dealing with a highly
engineered bioweapon. Since we are aware of this now, it would be
good to develop an IL-4 antibody so we can suppress this effect in vivo.
(Actually IL-4 antibodies probably already exist, we just need a humanized
version cleared for clinical use).
> It seems likely that by simply picking one or a few
> immune-affecting genes one would have a pathogen that did nasty things
> without having any obvious toxic effects when viewed as a genome - the
> effect is only due to the host reaction.
Yes, but its got to use host signal factors or mimics thereof to have
those effects on the host. Such genes are not normally found in
viruses or bacteria. (I'll note as an aside that viruses such as
SV40 and papilloma virus have managed to evolve their own genes
that do gum up the cellular works -- it would be useful to have
a complete catalog of these. We don't at this point -- but then
again neither do the terrorists).
> And there are plenty of conotoxins, few of which have been sequenced...
More than a few it looks like from scanning the nucleotide database.
But Blast and Smith-Waterman comparisons from a novel genome into the
nucleotide and protein databases are childs play at this point. If
the nasty bits are based on anything known we will know what they are
within a day.
> You seem to miss a step here. If I come up with a great in silico
> antitoxin, it still has to be tested in vitro before starting to brew it
> everywhere.
Yes. That is why it might be good to have pre-prepared and tested
molecules whose toxicity profiles are known. This information is
likely to already be in many PharmaCo databases -- one has to get
them to share it with the government though.
> Again, this takes a while, especially if it is not obvious
> how to synthetise it (how quickly can biochemists devise a workable
> synthesis pathway with no problematic residues?
We have retrosynthesis programs now to figure this out, I'm thinking
of getting one to wrestle with the Fine Motion Controller.
> and if it is protein based, don't you have to solve the folding problem
> to see if it sticks to the toxin?).
I'm assuming you have the crystal structure of the toxin and for
small peptides de novo folding works pretty well at this point.
Bule gene and/or Folding@Home will push our capabilities here
even further.
> Also, what about in vivo testing? In a crisis people might prefer to be
> the guinea pigs of untested antitoxins rather than a fairly certain
> death, so safety considerations might be ignored, but considerations of
> effectiveness are very relevant.
I'm assuming that is likely to be the case. Obviously if we are building
libraries of humanized antibodies we can screen out those that react
excessively with natural human proteins. That leaves us with a very
large pool to select from regarding binding strengths and silencing
of toxin activity. Here the safety concerns would be even less than
a new drug molecule based anti-toxin.
> It seems that the proper approach would
> be to try to have a number of antitoxins being developed and tested in
> parallel rather than betting all on a single card.
Yes absolutely, that is in my paper because Robiobotics can play a key
role in facilitating this.
> I think a project like this could be done, and it might be very useful
> for a huge crisis. But I think you are overly optimistic about how fast
> it could produce a stopgap measure even if the infrastructure was in
> place.
It depends on how much you have in place. I think a schedule would go
something like 1 day to isolate, 1 day to sequence, 1 day to identify
the bad actors, N days to discover/design prototype antitoxins,
X days to produce them, Y days to test, Z days to produce the
best candidates and get them to affected medical centers.
The trick is to get N, X, Y & Z as close to 1 as possible. I think
if you designed it on the basis of a high amount of parallelism
this could be done. You have to keep in mind that we have robots
now that manipulate 1500 or more samples simultaneously. There
are production lines out there now that are doing this 24/7.
Robert
This archive was generated by hypermail 2.1.5 : Sat Nov 02 2002 - 09:16:13 MST